One attribute that needs further highlighting is the work type attribute. This attribute is important because it determines the work location and the time of day that the agent travels to that location.
Census data includes very useful information on commuting patterns. By combining this with land use data it’s possible to include a work location that is based upon an agent’s work type. Therefore the commuting destinations within the simulation are more likely to reflect real-world travel patterns.
Home locations, work locations and other attributes are used to devise a travel plan for each agent. This is essentially an itinerary of activities that the agent travels to within the model. Each activity (such as home, work, shop, leisure) has a location and a provisional start and end time.
The chart below shows the daily activities for a small selection of agents.
Note the variation - each agent’s plan is unique and is calibrated to take account of factors such as school start times, work shift patterns and retail opening times.
The simulation is run over a series of iterations (100+). Each agent receives a score after each iteration. This score reflects how efficiently the agent performs their travel plan.
Agents can have different scoring parameters (depending on their attributes).
In subsequent iterations agents have the option of altering their travel plan, by choosing different travel modes, different routes or shifting activity start times.
Go to part 3 of the MATSim overview guide.